Versioned ConfigMaps

This release introduces versioned ConfigMaps. No more need for configmap-reload sidecar containers!

Two bug fixes - tcpinfo accounting, and etcd member count.

#270 use tcpinfo v1.1.0 with finwait2 bug fix 19 minutes ago
#269 Fixes the logic about how many etcd members are healthy 4 hours ago

Update ndt-server (new result schema)

Update NDT version (#267)

Deploy DISCO as core service, new recording-rules, and management updates

New script to easily add ndt-cloud platform cluster nodes (#256)
Automatically uses mlab4 machine for staging cloud VMs, and mlab1 for sandbox and production. (#257)
Modularizes containers in jsonnet templates file (#258) …
Mount node-exporter data directory for textfiles (#262)
Sources /root/.bashrc so that etcdctl commands work. (#264)
Add "utilization" experiment to daemonsets/core (aka DISCO, collectd) (#265)
Fix add_k8s_master_node.sh (#266)

Update NDT with new metrics

#254 Update ndt-server version to v0.11.0 - improved result & error metrics

tcpinfo tweaks, bismark on sandbox only

Merged #255 sandbox tweak-bins
Merged #253 Wraps all net aggregation rules in rate() and multiplies by 8 to get bits/s
Merged #252 A small fix for network aggregation Prometheus recording rules.
Merged #251 Adds new recording rules for aggregate network usages.
Merged #246 Refactors pretty much all CPU and memory Prometheus recording rules
Merged #250 Updates the GCE machine type for each project for the Prometheus VM
Merged #248 Add mlab/project label to nodes + configure bismark to only run in sandbox

TCP info new metrics

Merge pull request #249 from m-lab/sandbox-tcpinfo-metrics-2

Sandbox tcpinfo metrics for sidestream dashboard

Deploy download close bugfix

This release includes fixes to resolve:

• m-lab/ndt-server#160 server should close download connection after 10sec

Removes --before-reboot-annotations flag from CLUO

• The --before-reboot-annotations flag to the update-operator was removed because it's implementation cannot easily be managed manually.

Updates for CLUO + fixes to a couple Prom recording rules

• The Container Linux Update Operator deployment was modified. The Agent is now only deployed to platform nodes. Master nodes will rely on weekly automated reboots, one master node per day.

• Fixes two Prometheus recording rules that were referencing old label node, but now reference machine.